13

The Southern Journal of Philosophy (1992) Vol. XXX, No. 3

THE SUPERVENIENCE ARGUMENT

AGAINST MORAL REALISM

James DreierBrown University

I. IntroductionIn 1971, Simon Blackburn worked out an argument against moral realism appealing to the

supervenience of the moral realm on the natural realm.1 He has since revised the argument, in part to

take account of objections,2 but the basic structure remains intact. While commentators3 seem to

agree that the argument is not successful, they have not agreed upon what goes wrong. I believe this

is because no attempt has been made to see what happens when Blackburn's argument is addressed

to particular varieties of moral realism. As I see it, we must look to these various brands if we want

to understand just where the concept of supervenience can be usefully employed.

In the first part of this paper, I set up the argument and apply it to an array of realisms, arguing

that it is not generally persuasive as a refutation of these. In the second part, I provide an alternative

conception of the semantics of moral expressions, anti-realist in spirit,4 arguing by analogy that it is

an appropriate solution to the puzzles raised by the supervenience of the moral on the natural.

II. Blackburn's Argument

What is Blackburn's argument? Before I can give it any sort of precise rendering, I have to say

something about what supervenience is. Blackburn and commentators have used a number of

different formulations of the various relevant claims. Rather than adopt any of theirs, or inject my

own into an already somewhat confusing menagerie, I shall use those developed by Jaegwon Kim5.

I do so partly in the interest of clarity, and partly in the hope that the debate can begin to adopt a

uniform standard.

A. Formulations of Supervenience

14

Here are two kinds of relations that may hold between, for example, moral properties and natural

ones. They are, respectively, Weak Supervenience (WS) and Strong Supervenience (SS).6 Suppose

A and B are families of properties closed under Boolean operations. That is, if P and Q are

properties in a family, then so are the properties of being both P and Q, being either P or Q, not

being P, and so forth.

WS: B(∀ F in A)(∀x) [Fx → (∃ G in B)(Gx & (∀y)(Gy → Fy))]

SS: B(∀ F in A)(∀x) [Fx → (∃ G in B)(Gx & B(∀y)(Gy → Fy))]

We are, of course, interested in claims that the moral supervenes on the natural. So A is the family

of moral properties, and B is the family of natural properties. Then the two claims can be interpreted

informally as follows. WS says that necessarily, for each moral property instantiated by some

object, o, there is a natural property that o has, which the moral property always as a matter of fact

accompanies. SS says that necessarily, for each moral property instantiated by an object, o, there is

a natural property that o has, which the moral property always as a matter of necessity

accompanies. We can put it another way. WS is equivalent to the claim that necessarily, if two

individuals in a world are alike in all their natural properties, then they are alike in all their moral

properties. SS simply lifts the restriction that the two individuals be in the same world. The different

renderings I just gave of WS can be shown to be equivalent, on the plausible assumption that

necessarily everything has some natural property7; likewise for the renderings of SS.

The alert reader will want to know what kind of necessity is being asserted. The answer is

forthcoming.

B. The General Argument

Blackburn's argument against moral realism is roughly this. A certain supervenience claim, one

connecting the moral realm to the natural, is true. But another, stronger claim is false. The realist

cannot explain why the weaker connection should hold, given that the stronger one does not, while

the antirealist can easily explain this. So realism accrues a debt of explanation which it cannot

discharge.

Before narrowing in on the precise sorts of supervenience at work in the argument, let's look at

what sorts of connections do, intuitively, hold between the two kinds of properties. We can arrange

the options by focusing on WS and SS and two ways of understanding the boxes, that is, the

15

necessity operators, in each. First, the outer boxes, the ones that bind the entire formulas, should be

read as analytic necessity. I take this to be uncontroversial, at least among parties to the Blackburn

debate. The idea is that whatever kind of supervenience holds between the moral and the natural,

does so by virtue of the nature of moral expressions and moral thought. The supervenience that in

fact obtains between these two realms, whatever it is, cannot be denied by anyone who has thought

carefully about the matter and is not abusing the expressions in question.8 At any rate, nothing in

my analysis will turn on how we read the outer boxes.

Given that the outer box expresses analytic necessity, is WS true? Surely it is. For if it were false,

then there might be two people who were just alike in all natural respects but who differed in some

moral way. That seems very implausible, and it is precisely what Blackburn wants to say is ruled

out by our moral concepts.9

What about SS? I want to evaluate SS relative to two readings of the inner box. Let us first focus

on metaphysical necessity, and call the resulting interpreted sentence, "SS(metaphysical)". Is that

sentence true? We can borrow Hare's example10 of Francis, whom we assume to be a very good

man. We know from WS that anyone just like Francis in every naturalistic respect is also good, and

to the same degree. What we now want to know is whether any possible person who was just like

Francis in every naturalistic respect would be good, and to the same degree. And indeed it seems as

though he would be, and that anyone who denied that he would be would thereby exhibit a linguistic

or logical deficiency. This intuition is shared by Hare, and I think by most philosophers who have

thought about the matter.

But now consider SS(analytic), strong supervenience with the inner box interpreted as analytic

necessity. As I shall be understanding analyticity, a sentence is analytically true if and only if it is

logically entailed by the meaning of the expressions it contains. Along the same lines, a sentence

analytically implies another just in case the conditional whose antecedent is the first and whose

consequent is the second, is an analytically true sentence. Now it is SS(analytic), I believe, that

Blackburn and others have wanted to deny. For SS(analytic), together with the factual premise that

Francis is a good man, implies that there is some description of Francis couched in purely

naturalistic vocabulary which analytically implies that he is a good man. As Blackburn insists, many

of us are strongly inclined to doubt that there could be such a description.

16

For to tell which moral quality results from a given natural state means using

standards whose correctness cannot be shown by conceptual means alone. It means

moralizing, and bad people moralize badly, but need not be confused.11

If SS(analytic) were true, then for each moral predicate satisfied by an object, o, there would be a

naturalistic predicate which applies to o, the satisfaction of which analytically guarantees the

satisfaction of the moral predicate. But consider Francis, who instantiates the moral property of

being very good. Consider any naturalistic description of Francis, and ask whether that naturalistic

description analytically implies the sentence, "Francis is very good." The answer we are inclined to

give is, "No naturalistic description of Francis analytically implies the sentence, 'Francis is very

good.'" For many of us accept Hume's dictum that no "ought" is derivable from any "is".

So the plausible position that Blackburn would have us take with respect to the various kinds of

supervenience is that WS and SS(metaphysical) are true, and SS(analytic) is false. We are now

ready to state a precise rendering of Blackburn's argument. It has two premises. The first is that

SS(metaphysical) is true12 , and the second is that SS(analytic) is false. The reasoning is that the

conjunction of the premises needs an explanation, and the realist has none available. We'll have to

see why the conjunction is supposed to need an explanation, and whether the realist can in fact give

one, as we canvas the realist positions available.

I must add at this point that I am not entirely confident that I have correctly represented

Blackburn's argument as he intended it. The main reason is that I do not understand it perfectly.

The formulas that he actually uses are not SS(metaphysical) and SS(analytic), but rather these13:

(S) N((∃x)(Fx & G*x & (G*x U Fx)) → (y) (G*y → Fy))

(N) N(x) (G*x → Fx)

Blackburn says that the “F” denotes some moral property, the “G*” some maximal natural

property that intuitively sums up the combination on which the moral one supervenes, and “U” the

“underlying” relation which holds between a pair of properties just when the first underlies the

second. The problem for the realist, he says, is to explain why (S) is true, but the stronger (N) is

false.

There are some technical problems explaining exactly what goes into a "maximal" natural

property, and Blackburn recognizes them, and they are not my concern. But there are two other

confusing features of the formulas. The first is that it is not clear what kind of expressions "F" and

17

“G*” are supposed to be. They don't refer to specific moral and natural properties, one suspects,

or Blackburn surely would have said which ones they are (at least which moral one "F" refers to).

But if they are variables, they ought to be bound by second-order variables, and it is not easy to see

what the scope of the variables should be and which ones. But the most troubling question is what

the "underlies" relation is supposed to be. One cannot take (S) to be itself a definition of

"underlies" (in the sense of understanding the intent to be that whatever relation will fit the formula

is defined as the underlies relation) since it is far too indeterminate. The most natural way to

understand it is as the converse of the supervenience relation. But if that were right, there would be

no need to write out a whole formula, (S), for (S) is supposed to assert that F supervenes on G*.

I don't mean to take Blackburn to task here, but only to explain why I am using substitutes for his

formulas. My best attempt at reconstruction of his argument yields the argument I've given here. If I

am mistaken, then what follows is not directly a criticism of Blackburn's actual argument, but of

another argument that is at any rate closely related. Since my goal is to use the argument as I

understand it to motivate a certain way of understanding moral predicates, I shall not go to greater

lengths to map what Blackburn says precisely onto my scheme, though I hope I have at least come

close to capturing his intent.

III. Some Varieties of Moral Realism

The central question seems to be, How effective is this argument against moral realism? But that

question is too vague. We need to look at specific varieties of moral realism, and see how well the

argument fares against each. In this section I shall sketch three forms of moral realism and

comment on their status with respect to Blackburn's argument.

A. Moral Dualism

Let's start with moral dualism, the view that moral predicates express real, nonnatural properties.

G.E. Moore famously held such a view. He believed that there is a property, moral goodness, which

is not identical to any natural property. Where, according to Blackburn's argument, does such a

view go wrong? The dualist seems to have trouble with SS(metaphysical). For SS(metaphysical)

posits a necessary connection between moral and natural properties, and according to dualism these

18

properties are of distinct kinds. Whence this necessary connection between distinct existences, to

borrow from Hume? Moore himself wrote, "If a thing is good (in my sense), then that it is so

follows from the fact that it possesses certain natural intrinsic properties."14 Just what the

connection is, and in what sense moral properties "follow from" natural ones, Moore was never able

to explain. The dualist, then, seems to be saddled with what Blackburn calls "an opaque, isolated,

logical fact, for which no explanation can be proffered,"15 an extra law of metaphysics.

B. Reductive Naturalism

But not many realists are dualists these days. Moral realism has taken a naturalistic turn, with the

result that the victory over Moorean dualism rings hollow. The dialectic against the naturalist turns

out to be an utterly different one.

I shall take the work of Philippa Foot as an example of what we may call "reductive naturalism".16

Foot's idea is to get at moral goodness and rightness by way of an account of virtues, states of

character which a person needs to live a happy life. Right acts are those that flow from these states

of character. Whether a person is a good person follows strictly from the natural facts about the

person along with the meaning of "good". So Foot's position reduces moral facts to natural facts,

and moral properties to natural properties.17

Where does Blackburn expect to trip the reductive naturalist? It should be fairly obvious that there

is no problem squaring reductive naturalism with SS(metaphysical). Of course two things cannot

differ in their natural properties without differing in their moral properties, for something's moral

properties follow from its natural ones, the reductivist will say. Just as obviously, the reductivist

cannot explain why SS(analytic) is false. That is because she thinks it is not false. The reductive

naturalist thinks that SS(analytic) is true. In fact, SS(analytic) is close to being a statement of

reductive naturalism's central claim, that there are naturalistic descriptions the satisfaction of which

analytically guarantee the possession of certain moral properties.

I cannot see that Blackburn's argument is of any use against this form of realism. If the argument

is sound, of course, then reductive naturalism is false. But that is merely because one premise of the

argument is simply the denial of reductive naturalism. Taken as an argument against Foot, then, it is

only question-begging. What Blackburn would have to do to defeat a view like hers is to give some

19

reasons for supposing that SS(analytic) is false. As Foot has noted, the reasons so far advanced

have been rather off-hand and dubious.18

C. Nonreductive Naturalism

As dualists are a minority among moral realists, so what I have called "reductive naturalists" are a

minority these days among naturalists. As I read their positions, Peter Railton, Nicholas Sturgeon,

and David Brink19 are all moral naturalists who agree with Blackburn that an individual's moral

properties cannot be deduced from naturalistic claims about it. Yet all of them wish to identify moral

properties with natural ones. The task of moral philosophy, according to the nonreductive naturalist,

is to say which natural properties are identical with goodness, without insisting that claims about

possession of these properties follow analytically from naturalistic claims.20 If that seems

mysterious, it shouldn't. The idea is simple.

Consider the relation between the sentence "John drank some water" and the sentence "John

drank some H2O."21 If the first is true, the second must be. Any possible world in which John

drank some water is a world in which John drank some H2O. That's because water is H2O. As we

might say, the property of being water is identical with the property of being H2O. Or anyway, so

many philosophers are happy to say since Kripke explained rigid designation in Naming and

Necessity. The expressions "water" and "H2O" rigidly designate the same natural kind. Now

suppose something like that is true of the pair of expressions, "good" and "producing the greatest

happiness of the greatest number" (or some other naturalistic predicate). They designate the same

property in every world.

Of course, someone could assert "John drank some water" and reject "John drank some H2O"

without making any linguistic or logical error. Neither follows analytically from the other. And, for

all that is shown by the supervenience argument, nothing prevents the nonreductive naturalist from

suggesting the same thing about the sentences "Francis is a very good man" and "Francis is a Φman", where "Φ" is a naturalistic predicate.22

But is it really part of the meaning of "water" that it names a physical kind? Or could someone be

a competent user of the expression without knowing that everything just like some water in physical

20

respects is also water? It seems to me that it is part of the meaning. But suppose not. Surely we

could introduce a term that had this feature, that one would have to know that it named a physical

kind, but not necessarily which kind (in the sense of being able to specify it in a canonical way, as a

material scientist would) to be a competent user. Let us call "cosmite" whatever element

predominates in the constitution of Halley's Comet. We let "cosmite" be a rigid designator whose

reference is fixed by the description "the element that predominates in Halley's Comet." Now, you

and I know that being made of cosmite supervenes on something's physical properties, but we don't

know (at least I don't!) which properties, we don't know any relevant reduction sentences. If your

intuitions differ from mine regarding the "water" example, substitute the "cosmite" case for the

analogy with nonreductive naturalism.

So the nonreductive naturalist can explain why SS(metaphysical) is true, while SS(analytic) is

false. No two possible objects can differ in their natural properties without differing in their moral

properties, on this view, because moral properties are natural properties. Moral words are rigidly

coreferential23 with some naturalistic predicate. On the other hand, the moral evaluation "Francis is a

very good man" does not follow analytically from any naturalistic description of Francis, any more

than "John drank some water" follows analytically from any "H2O" sentence. So SS(analytic) is

false. The nonreductive naturalist has no trouble explaining the conjunction of SS(metaphysical)

with the denial of SS(analytic); indeed, that conjunction follows from the nonreductive naturalist's

theory.

To what extent does this nonreductive approach depend on a Kripkean theory of rigid

designation? It seems to me that it is a sufficient but not a necessary condition for the plausibility of

the nonreductive moral naturalist's reply that something like Kripkean rigid designation is a

plausible theory of moral predicates. If "good" and some naturalistic description rigidly express the

same property, then SS(metaphysical) is explained. But a nonKripkean explanation of synthetic

necessities like "Water is H2O" might well carry over to the moral example as well.

In "Supervenience Revisited", Blackburn wants to claim that the explanation I have attributed to

the nonreductive realist is bogus. More precisely, he seems to think that it is a kind of explanation

that gives the game away to the antirealist. Here is what he says:

21

[The combination of the failure of the stronger kind of supervenience with the

obtaining of the weaker kind] has now been explained purely by the structure of

beliefs which can coexist with competence [in the use of the relevant expressions].

There is indeed no further inference to a metaphysical conclusion about the status of

wateriness, because the explanation which, in the other cases, that inference helps to

provide, is here provided without it. To put it another way, we could say that in the

moral case as well, when we deal with analytically possible worlds, we are dealing

with beliefs we have about competence: in this case the belief that the competent

person will not flout supervenience. But this belief is only explained by the further,

anti-realist, nature of moralising. If moralising were depicting further, moral aspects

of reality, there would be no explanation of the conceptual constraint, and hence of

our belief about the shape of a competent morality.24

I think this reply is seriously confused. Taken literally, it implies that no realist could ever explain

any analytic connections among concepts. Imagine a challenge to a realist about bachelorhood and

marriage, asking her to explain the necessary connection between being a bachelor and being

unmarried. Why is it that anyone who claims, "Juan is a bachelor", but denies "Juan is unmarried,"

thereby shows that she is confused about the predicates in those sentences?25 The realist tries to

explain in the obvious way: the meaning of "bachelor" logically implies that anyone who satisfies it

also satisfies the predicate "unmarried", given the meaning of "unmarried". It is a fact about our

language, about the way we use words that explains the necessary connection between the two

sentences. But Blackburn replies, "Aha, you have given the game away to us antirealists about

bachelorhood. For any explanation that appeals to our linguistic practices, rather than to

metaphysical facts, commits one to antirealism." That would be silly—of course bachelorhood is a

real property, and of course the realist is entitled to appeal to semantic facts to explain analytic

connections.

Here is what I think the nonreductive naturalist should say. It is a part of the meaning of the word

"good" that it names a natural property, just as it is part of the meaning of the word "water" that it

names a physical kind. Anyone who thinks two things could be alike in their physical properties

while differing in whether they are made of water, reveals ignorance of the meaning of "water". That

is a fact about our linguistic practices. And anyone who thinks two things could be alike in their

natural properties while differing in their moral goodness, reveals ignorance of the meaning of

"good". That is a fact about our linguistic practices. But being made of water and being good are

real properties nonetheless.

22

D. Summary of the Dialectic

On the whole, Blackburn's argument does not fare well against the varieties of moral realism I

have sketched here. One of the premises of the argument, SS(metaphysical), is difficult for the

moral dualist to explain, and very hard to deny. But against the dualist, the second premise does no

work. If we want to explain what seems bizarre about moral dualism, we can simply point out that it

fails to explain what everyone admits, that the moral supervenes in some way on the natural.

Against forms of moral naturalism, the argument is even worse. For it merely begs the question

against the reductive naturalist, asserting without argument that SS(analytic), which is the crux of

her view, is false. And the nonreductive naturalist has no difficulty explaining how

SS(metaphysical) could be true while SS(analytic) is false, since these follow from the

nonreductivist's theory. Blackburn's objection that the kind of explanation the nonreductivist is

compelled to give for why SS(metaphysical) is a conceptual or analytic truth (that is, why the outer

box is read as analytic necessity) must commit us to antirealism, is confused. Since

SS(metaphysical) is itself an analytic truth, of course it must be explained by reference to the way

we think and talk about morality rather than by reference to the moral and natural facts themselves.

IV. An Alternative Proposal

I will not speak to the plausibility of the kinds of realism I've discussed except insofar as I have

already done so. Instead, I want to look at an alternative view, one which is suggested to me by an

analogy between the kinds of supervenience which hold in the moral case and the way they hold, on

the one hand, and on the other the ways of supervenience of another familiar class of natural

language expressions, those called "indexicals".

A. Preliminary Note

To set the stage, I want to discharge a simplifying assumption. The supervenience of which we

have been speaking is a relation between classes of properties. But I now want to ascend to the

formal mode, and speak instead of classes of predicates and their supervenience relations. We will

need to introduce no new formulas at all, only to reinterpret the quantifiers as substitutional rather

than objectual.

There are a number reasons for shifting our footing. The first is that to speak of moral properties

23

at all seems to beg the very question at issue in favor of the realist. It is somewhat surprising that

Blackburn should use as a major premise in his chief argument against moral realism a claim that

the class of moral properties is related to another class of properties by supervenience, since this

claim seems to entail that there are some moral properties, and thus the realist has won before the

argument gets going. Could the point be to show that the class is empty? Perhaps, but Blackburn

never draws that conclusion. Is it that moral properties are mere quasi-properties, and not real

ones?26 That sounds like something Blackburn might say, but it is hard for me to interpret the

quasi-real/real distinction in such a way. It seems to me that the natural way to understand the

quasi-realist program is to take it as saying that the correct semantics for moral expressions licences

(almost) all realist-sounding talk, but is not realist precisely in that it does not assign any properties

as the values of moral predicates. If we run the argument at the level of predicates, we at least leave

open the possibility of a conclusion that does not contradict the premises.

Second, reframing the argument in the formal mode heads off a potential objection regarding a

crucial condition, the closure condition. Remember that the statement of the supervenience claims

requires that classes be closed under Boolean operation. But there is some question whether classes

of moral properties, or any properties, can survive closure.27 There is no question, though whether

classes of predicates can be defined to be closed under Boolean operation, or if there is, the obstacle

is easily skirted; we simply define a class of moral formulas as the closure of the class of moral

predicates (we won't need infinite formulas, though there is no logical objection to including them).

Third, by our linguistic ascent we eliminate the difficult problem of specifying what we mean by

"natural" properties. Moore, for whom such a specification was indispensable, had tremendous

difficulty delivering one. Instead, we need only specify which classes of predicates are to supervene

on which, and this may be done simply as follows: the supervening class is the closure of the class

of moral predicates, and the supervenience base is the closure of the class of scientific predicates.

We may include all predicates in the language of the sciences, including psychology and the social

sciences, in the latter class, and the former class is, I think fairly intuitive–if not, we may restrict it

(for the purposes of the present argument) to the closure of the class containing "good", "bad",

"right", and "wrong".28

24

The fourth reason is a technical one. When we read the inner box of SS as analytic, we are not

taking it to be a propositional operator (at least this is so as I have explained analytic necessity).

Rather, the box operates directly on the sentence. Thus, SS(analytic) is sensitive to which predicates

are substituted for variables when we instantiate them. Since I will be instantiating with examples

below, it is best to use substitutional quantification and acknowledge that we are talking about

predicates and their relations, rather than about properties. And in general, I will be suggesting that

it is wiser to speak of moral predicates instead of properties, for a reason related to the first reason,

above.

B. An AnalogyPretend that all of us are confused, philosophically perplexed, about indexical expressions. We

know how to use these words, and confidently assert such sentences as "I arrived in New Orleans

yesterday," but when it comes to giving an analysis of "yesterday" we feel at a loss. And suppose

someone suggests a kind of realism about such expressions. He says that there is a certain

property, yesterday occurrence, a property of events, which some have and some do not have,

independent of our evidence, and so forth.

It seems a plausible enough view in certain respects. For example, we do believe that when

someone claims that the invasion occurred yesterday, and someone else claims it did not occur

yesterday, we of course believe that one of them must be mistaken. And it does seem to us that

whether or not a given event occurred yesterday is not dependent on anything about us.

But now comes Simon Blackburn, who alerts us to what he claims are troubling facts about

yesterday occurrence. In particular, he asks us to look carefully at a supervenience base for

temporally indexical predicates: namely, what we might call the "date-like" predicates. These are

such ordinary beasts as, the predicate "took place on April 5th , 1960", the predicate "occurred on

April 25th , 1990", and the like. There is, it seems, a family of predicates of the form occurred on

_________. It would be a trivial and excruciatingly boring enterprise to give the generative rules for

these predicates.

Now Blackburn urges that we investigate the supervenience of yesterday occurrence on the family

of date-like predicates, and so we shall.

Let A be the family of the predicate of yesterday occurrence along with all of its Boolean

25

permutations (there aren't many; just the predicate and its complement). And let B be the family of

date-like predicates. What kind of supervenience will hold between them? WS certainly seems to.

For any pair of events which satisfy all the same date-like predicates (that is, they occur on the same

date!) must agree in whether they occurred yesterday.29 Quite so; fixing the date on which an event

occurs fixes its yesterday occurrence in at least this way!

What of SS(metaphysical)? Intuitively, if the world's best performance of Mozart's Requiem

occurred on a certain date and also occurred yesterday, then any possible event which had occurred

on that date would have occurred yesterday. So the supervenience holds across metaphysically

possible worlds, too, which is after all what SS(metaphysical) states.

And SS(analytic)? This is the tricky one. The question is whether, given that a certain event

occurred yesterday and also occurred on April 25th , 1990, it is analytically necessary that any event

which occurred on April 25th , 1990 also occurred yesterday. But in fact it does not seem to be

analytically necessary! Fans of John Perry's "The Problem of the Essential Indexical"30 , or

Richard's "Direct Reference and Ascriptions of Belief"31 , will easily be able to give examples

demonstrating why not. For suppose that the Celtics' worst loss of the season occurred yesterday,

and it occurred on April 25th . And suppose we ask Rip van Winkle, who is an avid Knickerbocker

fan, but has no idea what day it is, whether the Celtics' worst loss occurred yesterday. He's just seen

in a newspaper that their worst loss occurred on April 25th , but he is unable to conclude by mere

logic and conceptual analysis whether April 25th was yesterday. The knowledge he lacks is not

logical or conceptual. It is not, after all, an analytic truth that yesterday was April 25th . So

SS(analytic) does not hold between the two families. Yesterday occurrence does not strongly

analytically supervene on date-like predicates.

Blackburn now asserts that the combination of the holding of SS(metaphysical) and failure of

SS(analytic) between indexical and date-like predicates poses a puzzle for a realist about the former.

For what could explain the weaker connection without entailing the stronger? Claims about what

occurred yesterday must have a kind of emotive function, expressing a yesterdayish attitude about

them rather than attributing a genuine property. But I trust that by now we will not be taken in!

What are our options?

First, we could be dualists. Perhaps yesterday occurrence expresses a property, but a peculiar,

26

non-date-like sort. But no, this would leave it quite mysterious why yesterday occurrence is

connected by SS(metaphysical) as they are to date-like predicates.

Second, we could be "reductive datists": yesterday occurrence does, we might try to maintain,

follow from sentences about what day it is (or was). But the view flies in the face of the Rip van

Winkle example.

Third, we could consider a nonreductive identity view. Perhaps "yesterday" is the name of some

particular day, but not analytically equivalent to any date name? Enough is enough; it is difficult to

take these accounts seriously! Neither anti-realism nor any of the straightforwardly realisms are in

the least plausible accounts. Let's see how the correct account of yesterday occurrence fits into the

supervenience argument.

I shall first rehearse in broad outline the familiar Kaplan account of indexicals.32 On this view,

indexicals have as their primary semantic value a character, a rule by which their content is

determined on an occasion of use. For the indexical "I", the rule is simply that the referent is the

speaker, and since indexicals are directly referential, the referent is the content. For the indexical

"yesterday", the rule is that the content is the day prior to the date of use. By extension, for

occurred yesterday, the rule is that the content is the property of having taken place the day before

the date of use. These rules are the characters of the indexical expressions, and they constitute their

primary meaning in the sense that all and only competent users of the expressions know these

rules.33

Since this account is the correct one (so I believe), it must deliver the intuitively correct results

when plugged into the various supervenience claims, and of course it does. Take WS first. Is it

necessarily true that for each yesterday occurrence (or YO) predicate there is some predicate such

that every event that satisfies the date predicate also has the property expressed by the YO

predicate? Of course; or at least it is once we fix an occasion of use. For example, if the predicate is

occurred yesterday (OY) itself, and the day of use is d, then the relevant date property is the

property of having occurred on d-1. And similarly for SS(metaphysical). To put it formally, the

Kaplan semantics allows us to prove that for any occasion of use,

(∀x) [OYx → (∃ Date, D)(Dx & B(∀y)(Dy → OYy))]

and

(∀x) [OYx → (∃ Date, D)(Dx & B(∀y)(Dy → –OYy))]

27

where the box is read as metaphysical necessity. Since OY and –OY are the only members of A,

this suffices to prove SS(metaphysical). And the idea is intuitively simple.34

We also want to show that SS(analytic) is false on the Kaplan model. Again, the idea is fairly

straightforward. On a given date of use, we cannot find a date expression which is analytically

equivalent to OY, for all someone has to know in order to know what OY means is its character, the

rule for determining its content. He does not have to know the content itself. And, the rule alone

does not yield a corresponding date predicate. That's why Rip van Winkle could know English

perfectly well and not know that yesterday was April 25th .

All of this suggests to me an analogy between moral expressions and indexicals. Suppose we

think of moral predicates as expressing properties relative to a context, much as indexicals like "I"

and "here" refer to persons and places relative to a context of use. There is a rule associated with

"morally good"; its character, which determines a natural property expressed for each context of

use. The rule could be to find the natural properties endorsed by the community of the speaker, or

of the agent being judged, and then we have kinds of cultural relativism. I have elsewhere spelled

out more fully what such an indexical account would look like35, and given an argument for why it

is independently plausible. Here my aim to to display another virtue of such an account: that it can

explain the sorts of supervenience relations that do obtain between the moral and the natural realms.

Let me now fill in at least a rough outline of how an indexical account would work.

Here is a simple, and too crude, variety of an indexical sort of moral relativism. It says that each

speaker has what we may call a "moral system," comprising the sorts of moral attitudes and

affective states which anti-realists generally say exhaust the semantic content of moral utterances.

When a person with a moral system, M, says "x is morally good," according to this view, she is

asserting that x has a certain natural property, P. Which natural property? P is the property of being

rated highly by M. It follows, of course, that when different speakers say "x is morally good" they

may be asserting of x that it has different natural properties, each determined by the speaker's own

moral system. The content of "x is morally good" varies, then with the context of utterance, and

more particularly on the speaker in that context and on the speaker's moral system.

Now how would supervenience claims turn out on such a view? They turn out just as in the

28

"yesterday" example. WS comes up true, because on a given occasion of use a moral predicate

expresses a natural property, so on each occasion there is some natural property which can be truly

said to accompany "goodness" invariably. The parallel to the "yesterday" example, wherein on each

occasion some date could be truly said to be coextensive with "yesterday", should be apparent. And

similar remarks again apply to SS(metaphysical). Just as "yesterday" (when used on April 26th) is

necessarily coextensive with "April 25th", so "good" when used in a certain context would be

necessarily coextensive with whatever predicate picks out the natural property "good" expresses in

that context.36

What about SS(analytic)? It turns out false on the present model, as it does intuitively. From no

naturalistic description of something will it follow analytically that it is, say, good, just as from no

date-like description of a day will it follow that the day was yesterday. Knowledge of the character

of "good" does not alone suffice to produce knowledge of which properties are, in the relevant

context, the "good making" ones. A complete knowledge of all features of the context would be

required for that.

C. Some Salient Features of the Analogy

It turns out that what we might call an indexical theory of moral expressions fits our

supervenience intuitions perfectly. Of course, so does nonreductive naturalism, and so does

Blackburn's sophisticated emotivism. I cannot analyze all of the merits and drawbacks of the three

views here, but I will say something about why an alternative should be welcome.

Roughly, I think the realist views so far canvassed face the following difficulties. Dualism is

correctly accused of positing supervenience as brute, unexplained. Reductive naturalism seems to

most of us to be too ambitious; our deeper moral disagreements are simply not going to be

dissipated by any amount of conceptual analysis. Nonreductive naturalism has not so far (and

prospects are dim) been able to provide a satisfactory account of how moral expressions might

come to refer to the particular natural properties they do refer to on any given naturalist account.37

Blackburn's Quasi-realism, of course, suffers from none of these failings. But, though it has come

some distance from Ayer's emotivism in addressing Geach's problem,38 noncognitivism still has

difficulty accounting for the apparently cognitive role that moral judgments play in ordinary

29

contexts more complex than simple assertions of right and wrong.39

None of these problems amounts to anything like a knock-down refutation, but I think they are

serious enough to motivate consideration of available alternatives. So I want to end by showing how

the indexical alternative can help us with two general issues that pervade the current meta-ethical

debate. The first is what is sometimes called the ineliminability of normativity. The thesis has a

fancy name, but all I mean is the apparent fact that normative concepts seem to inhabit their own

realm of language and thought. Any theory which purported to be able to replace all normative

vocabulary with pure description is these days met with skepticism, and rightly so, I think. What are

we to make of this fact? I believe we may turn to recent work in the theory of indexicals for an

answer.

In his well-known "The Problem of the Essential Indexical", John Perry shows that it is not even

in principle possible to replace all indexical expressions in a language with nonindexical ones.40

Roughly, the idea is that in such a rarefied language it would be impossible to pose the questions

"Where am I?" and "Who am I?" in the proper way. But Perry, and more recently David Kaplan,

argue that ineliminability does not mean, as Frege thought, that indexicals express some special

essence, some concept beyond the natural world. Rather, indexicals have semantic value in a

different category from other expressions, namely, a character but no fixed content. It is this

semantic property which generates the mysterious quality of first person attributions and their kin.

Now the indexical account of moral concepts offers to dissolve the mystery of the ineliminability

of the normative in a similar way; more than similar, in fact, since the account has it that moral

concepts are, in effect, indexical. The reason we could never satisfactorily replace moral vocabulary

with pure descriptions is the same as the reason we could never replace indexical vocabulary with

nonindexicals. The replacing expressions could not possibly have the right kind of semantic value.

This feature of the analogy, and indeed my proposal in general, raises the question whether the

indexical view is simply another version of a well-known whipping horse of meta-ethics,

Subjectivism. The Subjectivist takes moral claims to be claims about or expressive of the speaker’s

attitudes. It is widely thought to have been refuted long ago by some arguments of Moore’s and

30

Blandshard’s.41 The arguments are semantical and purport to show that the Subjectivist analysis

cannot possibly be a correct explanation of what we actually mean when we use moral expressions.

Myself, I have always felt that they are a kind of trick, that they exploit incidental features of the

particular analysis proffered but do not cut to the kernel of subjectivism. I suspect that insofar as

Moore and Blandshard won their debate with Subjectivism, it was because they were the cleverer

analysts and not because they had identified anything deep about ethics that showed the opponent

to be seriously misguided. In fact, I think that the indexical theory of moral language is the inheritor

of Subjectivism, and that it has the resources to answer the objections put to its ancestor by

objectivists like Moore and Blandshard. Of course, whatever really deep problems Subjectivism

suffered from will plague the indexical view as well. I believe that we must live with these problems,

whatever they may be. But for now, let’s look briefly at the old objections and see how the indexical

theory answers them. In so doing we will better understand the nature of the indexical analysis.

I will mention two kinds of anti-Subjectivist arguments which are fairly typical of the semantic

sort that were familiar and widely known by the middle of this century. The first related pair I’ll call

the “Contradiction Problem” and the “Disagreement Problem,” and they are due to Moore.42

Moore shows that according to Subjectivism, the same action might be both right and wrong at the

same time, because one person might have the relevant feeling of approval toward it while another

might have the relevant feeling of disapproval. The problem can be sharpened and made to look like

a strict reductio if we change the example slightly so that the same action comes out both wrong

and not wrong. Perhaps Subjectivism did have this consequence. But the indexical view clearly does

not. According to the indexical view, one person might say “Action A is wrong” and another say

“Action A is not wrong,” and both utterances might be true. It would be a mistake, however, to

think that a contradiction can be derived. The mistake would be in applying disquotation to

indexical utterances. Consider the following fallacious argument:

(1) Pierre said, “The Eiffel Tower is here,” while standing under the Tower.

(2) Sally said, “The Eiffel Tower is not here” while standing in Times Square.

(3) What Pierre said is true, and what Sally said is true. [(1) and (2)]

(4) “The Eiffel Tower is here” is true. [Substitution]

(5) The Eiffel Tower is here. [Disquotation]

31

(6) “The Eiffel Tower is not here” is true. [Substitution]

(7) The Eiffel Tower is not here. [Disquotation]

(8) The Eiffel Tower is both here and not here. [Conjunction]

Someone might try to use this argument to prove that a “Subjectivist” theory of ‘here’ is

incoherent. But of course, the reply is that the use of disquotation is illicit when the quoted sentence

contains an indexical. So Moore’s argument does not refute the indexical view.

But it does raise a second related problem, the Disagreement Problem. Moore pointed out that

according to Subjectivism, if one person says “Action A is right” and another says “Action A is

not right,” they are not really disagreeing, because each is only reporting about her own attitude.

Since we do take pairs of claims like that to show a genuine disagreement, Subjectivism must be

false. And here the problem seems to apply equally well to the indexical view. I have argued

elsewhere43 that this problem can be handled by the indexical view. In brief, the reply is that in

contexts like the one imagined, two speakers who we think are disagreeing are assuming that there

is a substantial area of overlap between their moral systems. The situation, I believe, is parallel to

one in which you and I are together looking through a pile of papers looking for an important

document. You say, “It just isn’t here,” and I say, “Yes, it must be here.” If we took a crude view

of ‘here’, thinking it referred directly to the place of the speaker, we’d have to admit that we were

not really disagreeing. But in fact, our uses of the indexical ‘here’ should be taken to refer to a

common area of overlap between us. Just so with the moral indexicals, or so I would argue.

Blandshard gave an argument relying on what he called “the method of difference.”44 He

imagines that he has come across a rabbit suffering in a trap, in great pain, and that he says, “It is a

bad thing that this animal is suffering.” He asked himself whether he would say that it would not

have been bad if he had not happened by, and naturally answers that he would judge that it would

have been just as bad. But, he notes, according to Subjectivism, his judgment turns out to be false

when evaluated in the circumstance in which he is ignorant of the suffering, because in that

circumstance he would have no attitude whatsoever toward the rabbit’s plight. This objection does

seem to show that Subjectivism fails to capture what is meant by the claim, “It is a bad thing that

this animal is suffering.” Notice, however, that the indexical view has a subtly but significantly

different analysis of the claim. According to the indexical view, ‘bad’ picks out a natural property

32

on each occasion of use. The content of the sentence, “It is bad that this animal is suffering,” is

that the event of the animal suffering has the natural property picked out by ‘bad’ in the context.

Presumably the property is something like being an instance of innocent suffering. Would that

event have had that property even if Blandshard had not happened by? Of course. So Blandshard

could truly say, “It would have been bad even if I had not know about it.”

Compare the indexical account of ‘here’. A crude theory, the analogue of the crude Subjectivist

view of moral expressions, might hold that the word ‘here’ simply means “the place I am now

speaking.” If that were correct, then the following would have to be true: “If I were on the moon,

there would be no air here.” But we ordinarily take such a claim to be false. If I happened to be on

the moon, that certainly would not change the atmosphere here on earth. The mistake is to confuse

the character of ‘here’ with its content.45 Perhaps that was the mistake of early Subjectivists with

respect to the meaning of moral expressions.

So if the question is whether the indexical view is in the spirit of the old Subjectivism, then the

answer is that it is. I think, though, that the indexical view is well suited to answer the semantic

objections that plagued Subjectivism.

I said above that the indexical view helps us to understand the ineliminability of normativity. My

discussion of that issue led to a comparison of the present view with Subjectivism. And this

comparison in turn introduces the second major issue on which the indexical theory can shed some

light, the issue of moral realism. Since Blackburn's original argument was supposed to be a

refutation of moral realism, one naturally wants to know where the indexical account stands with

respect to realism. The answer is not perfectly clear, mostly because of the unclarity of the terms of

the realist/antirealist debate.

In "Moral Realism", Peter Railton asks the following questions designed to separate realists from

antirealists in ethics:

Cognitivism–Are moral judgments capable of truth and falsity? .... Objectivity–In

what ways, if any, does the existence of moral properties depend upon the actual or

possible states of mind of intelligent beings? Reductionism–Are moral properties

reducible to, or do they in some weaker sense supervene upon, nonmoral

properties? Naturalism–Are moral properties natural properties?....46

His own view is supposed to fall on the realist side of each feature, thus he calls it "stark, raving

33

moral realism." An emotivist account is clearly antirealist, and Blackburn's view is mostly that

(though he wants to call it "quasi-realism"). But the indexical account straddles the fence. Whether

one wants to call it ultimately realist or not will depend on which features of realism one thinks are

most important in ethics.

Thus, on the indexical view any moral utterance is plainly capable of being true or false, and

expresses a proposition; it has at least this much realism about it. Further, the proposition expressed

will in general be true or false independent of anyone's evidence or mental states. That's because the

proposition expressed will be quite an ordinary sort of proposition, one which could be stated in

perfectly naturalistic vocabulary. Remember that it is not the proposition expressed that is especially

moral, on the indexical view, but the way it is expressed, just as the proposition expressed by

"Yesterday was the 25th" expresses no peculiar sort of fact but only a very ordinary date-like fact, in

a special indexical way.

On the other hand, there is something in the spirit of realism that is surely missing from the

indexical account. For a self-styled realist will usually want to claim that there is a single correct

moral scheme of things, and that conflicting moral claims should be understood as competing

claims about that single scheme. The indexical account offers no place for the robust realist's single

correct scheme, though, and so cannot be counted as "raving realism".

The point can be made nicely, I think, by asking the watershed question, "Are there any moral

facts?" All of the realists I mentioned above say that there are. Blackburn, Hare, and their ilk say

that there are not. But the indexical account can say only that there seems to be something wrong

with the question. Think of what we would say if someone wanted to know whether there are any

"yesterday" facts. The answer is that there are no facts that have yesterdayness built in, as it were–

which are the yesterday-laden facts will change from day to day! But there are, of course, perfectly

good facts that can be expressed by such sentences as "The Celtics lost yesterday." So the indexical

theory of morality replies that there are no especially moral facts. There are perfectly good facts

expressed by sentences such as "Invading one's neighbors is wrong," but these facts have no moral

stain about them. The morally laden facts may change from context to context.

A similar observation is made by James Klagge.47 Klagge distinguishes what he calls "ascriptive

supervenience" from "ontological supervenience", arguing that the antirealist is entitled to the

34

former and that the realist needs to explain the latter. Ascriptive supervenience is a conceptual

constraint placed by moralizers on moralizing, while (according to Klagge) ontological

supervenience is a real relation between classes of properties in the world. Myself, I think that all

supervenience claims (in ethics) are conceptual ones; the first operator is going to be most plausibly

read as analytic necessity. Notice that although the supervenience of yesterday occurrence on date-

like predicates hold in every context, it does not hold across contexts, and similarly, if we follow my

suggestion, moral predicates will supervene on natural ones within each context but not across

them. Klagge proposes that we need a new kind of supervenience, ascriptive supervenience, to

capture this idea, while I maintain that we can make do with the sorts of supervenience we have

already. But, the present suggestion follows the spirit of Klagge's distinction. The explanation of

supervenience I propose would tie all judgments made from a given perspective to some definite

subvening base, but would not assert an analytic connection between any perspectiveless

(extracontextual) moral judgment and any particular natural properties or facts. For moral

judgments, on the present proposal, always state natural facts (that is, their content is always some

proposition attributing a natural property to an individual or class), but the same judgment (in the

sense of "same sentence") may express different natural facts from different perspectives.

These remarks on the question of whether and in what sense there are moral facts relies rather

heavily on Kaplan's model of how indexicals work. I am taking for granted, for example, that there

are no indexical facts, no facts which have themselves any sort of indexicality built in. There are,

rather, indexical sentences which can express ordinary sorts of facts. But though I do favor

Kaplan's model, I (and he) might be wrong. Perhaps there are, after all, special indexical sorts of

facts. Perhaps some propositions, and not just sentences, are essentially indexical, as Sosa has

suggested. If so, then the precise ways in which the indexical view of moral expressions turns out to

be realist or anti-realist may shift. The supervenience relations between moral terms and natural

ones wouldn't change, nor the general form of the explanation of these relations, but the underlying

metaphysics of indexicality could be quite different. But in any case, the indexical view of moral

expressions helps us locate the spots where the partial truths of realism and anti-realism in ethics

may be located.

So several features of the indexical account seem to me attractive. For one, it does a neat job of

35

capturing our intuitions about supervenience. For another, it shines light into the fog of the

irreducibility of normativity. Finally, it explains what many of us have begun to suspect, that the

realist/antirealist debate has been framed in such a way as to force meta-ethical views into an

unnecessary dichotomy. For these reasons, I recommend consideration of the view as an alternative

to the realism against which Blackburn's argument was designed and to Blackburn's own antirealist

account. The supervenience argument cannot, as I think I have shown, do the dialectic work

Blackburn claims for it. But I think it does help us to get some of the positions more clear, and I

believe it also performs the valuable service of leading us toward a suggestive parallel between

morality and indexicality.*

References[1] Blackburn, Simon. 1971. "Moral Realism," in John Casey, Morality and Moral Reasoning

(London: Methuen).

[2] -------------. 1984. Spreading the Word (Oxford: Oxford University Press).

[3] -------------. 1985. "Supervenience Revisited," in Ian Hacking, ed., Exercises in Analysis:Essays by Students of Casimir Lewy (Cambridge: Cambridge University Press)pp. 47-67.

[4] Blandshard, Brand. 1949. “The New Subjectivism in Ethics,” Philosophy and Phenomeno-logical Research 9 pp. 504-11.

[5] Brink, David O. 1986. "Externalist Moral Realism," Southern Journal of Philosophy 24(supplement) pp.23-42.

[6] Dreier, James. 1990. "Internalism and Speaker Relativism," Ethics 101 pp.6-26.

[7] Foot, Philippa. 1978. Virtues and Vices (Berkeley: University of California Press).

[8] Geach, Peter. 1960. "Ascriptivism," Philosophical Review 69 pp.221-5.

[9] 1965. "Assertion," Philosophical Review 74 pp.449-65.

[10] Hale, Bob. 1990. "Can There Be a Logic of Attitudes?" in Haldane, John, ed., Reality:Representation and Projection (London: Oxford University Press, in press).

[11] Hare, Richard M. 1958. The Language of Morals (New York: Oxford University Press).

[12] . 1984. "Supervenience," Proceedings of the Aristotelian Society (58 supp.) pp.1-16.

[13] Horgan, Terence, and Timmons, Mark. 1990. "Troubles on Moral Twin Earth: MoralQueerness Revived," Synthese 92 (1992) 221-60).

[14] Kaplan, David. 1983 "The Logic of Demonstratives," Journal of Phil. Logic 12 pp.425-52.

[15] ----------. 1989. "Demonstratives," in Themes from Kaplan (Oxford: OUP) pp. 481-564.

[16] Kim, Jaegwon. 1984."Concepts of Supervenience," Philosophy and PhenomenologicalResearch 45 pp.153-76.

[17] Klagge, J. 1984. "An Alleged Difficulty Concerning Moral Properties," Mind 93 pp.370-80.

[18] ----------. 1987 "Supervenience: Perspectives versus Possible Worlds," PhilosophicalQuarterly 37 pp.312-15.

36

[19] ----------. 1988. "Supervenience: Ontological and Ascriptive," Australasian Journal ofPhilosophy 66 pp.461-9.

[20] Lewis, David K. 1979. "Attitudes De Dicto and De Se," Philosophical Review 86 pp. 513-43.

[21] McFetridge, I. G. 1985. "Supervenience, Realism, Necessity," Phil. Quarterly 35 pp.245-58.

[22] Moore, G. E. 1912. Ethics (Oxford: Oxford University Press).

[23] ----------. 1968 "A Reply to My Critics," in P. Schilpp, ed., The Philosophy of G.E. Moore(LaSalle, IL: Open Court).

[24] Perry, John . 1977. "Frege on Demonstratives," Philosophical Review 86 pp.474-97.

[25] ----------. 1979. "The Problem of the Essential Indexical," Nous 13 pp.3-21.

[26] Post, John. 1984. "Comment on Teller," Southern J. of Philosophy 22 (supp.) pp.163-7.

[27] Railton, Peter. 1986. "Moral Realism," Philosophical Review 95 pp.163-207.

[28] Sturgeon, Nicholas. 1985. "Moral Explanations," in David Copp and David Zimmerman, eds.,Morality, Reason, and Truth (Totowa, N.J.: Rowman Allenheld) pp. 49-78.

[29] Van Cleve, James. 1990. "Supervenience and Closure," Phil. Studies 58 pp.225-38.

Notes

1Blackburn, [1].2See Blackburn, [2] chapter 6, and [3].3The comments I am thinking of, and to which I will from time to time refer, are those given in Hare, [12], Klagge,[17], [18], and [19], and McFetridge, [21].4But see the last section for a caveat about what counts as "anti-realist."5Kim, [16].6These formulas are not equivalent to the ones used by Blackburn, McFetridge, Klagge, or Hare, but they are related.The arguments I will rely on can be easily reconstructed using the alternative formulations, most straightforwardlywith those given by McFetridge. All of the distinctions drawn by Blackburn can be drawn using Kim's apparatus.7Whether the assumption is plausible of course depends on what is meant by a “natural” property. Ultimately, Imean to include in the class of natural properties all those properties which are not moral properties.8Thus, Blackburn writes in Spreading the Word: "It seems conceptually impossible to suppose that if two things areidentical in every other respect, one is better than the other." (p.183) McFetridge agrees: "Our ordinary moral thoughtcommits us to (XXWA)", which is a kind of weakened version of our SS (p.249).9See previous note.10Hare, [11], p.153.11Blackburn, [2], p.184.12Hare, who Blackburn seems to think shares his position on supervenience, sometimes says that he wants to acceptWS and reject SS in general (Hare, [12], p.4). In fact, my evaluation of Blackburn's argument would apply equally toa version which used WS as its first premise. But I am quite certain that both Blackburn and Hare are committed tothe stronger SS(metaphysical)—stronger in the sense that SS(metaphysical) strictly implies, but is not implied by,WS. It cannot but strengthen the argument to use the stronger premise, which I take to be uncontroversial anyway.13Blackburn, [3], p.49.

37

14Moore, [23], p.588.15Blackburn, [1], p.111.16I am afraid that Foot would not be pleased at being called a reductive anything, but I hope it will be clear what Imean.17It hardly needs to be said that this thumbnail sketch of Foot's view does not begin to do justice to her full account,which is far too complex for me to analyze. Indeed, part of the point of Foot's work is that those philosophers, likeBentham, whom G.E. Moore accused of committing the naturalistic fallacy were guilty rather of having anoversimplified account of moral goodness. See Foot, [7], for a fairly detailed exposition of Foot's naturalistic ethics.18See especially her "Moral Arguments" and "Moral Beliefs" in [7].19For example, Brink, [5]; Railton, [27]; Sturgeon, [28].20I am going to use "nonreductive naturalism" as a name for the class of theories that take that task to be the centralone. It is in some sense a reductive task, but it does not include the reduction of meanings of moral expressions topurportedly synonymous naturalistic ones. If the reader prefers "nonanalytic naturalism" as a characterization of theclass I am interested, that's fine. I do not insist on the name.21McFetridge uses this example to draw a similar conclusion.22The pure at heart may prefer Quinean corner brackets around the second sentence, thus:

OFrancis is a Φ man.P23Can a predicate be said to be referential at all? Perhaps "rigidly coextensive" would be a better way of expressingthe idea.24Blackburn, [3], p.64.25Or is not serious, or is the world's worst logician.26I thank Nick Zangwill for suggesting this possibility.27See, for example, Post, [26], and Van Cleve, [29].28It is probably best to confine the class to "good" modified by degrees, thus,

Ogood to the nth degreeP for each n.29If you like, think of it the other way: Necessarily, for each yesterday occurrence predicate and every event, if theevent satisfies the predicate then there is some date-like predicate it also satisfies such that every event whichsatisfies the date-like predicate also satisfies the yesterday occurrence predicate.30Perry, [25]; see also his [24].31Kaplan, [14].32Kaplan, [15].33As Kaplan and Perry have argued, characters are not the same as Fregean senses, though they share importantfeatures. See Perry, [24], and Kaplan, [15].34The idea is to show that the sentences are true in any context, that is, on any occasion of use. If we can prove thisfrom the semantics alone, then we have shown that the sentence is analytically true. So consider

(∀x) [OYx → (∃ Date, D)(Dx & B(∀y)(Dy → OYy))]

We choose an arbitrary context, and call the day in that context d, and instantiate x with an arbitrary event, e.Suppose that e did not occur on the day before d. Then the conditional is true. Suppose that e did occur on the daybefore d. Then choose "Occurred on the day before d" to instantiate the predicate variable D. The first conjunct istrue, since e did occur on the day before d. And, in any world, every event which satisfies "Occurred on the day befored" also falls into the extension of the value of the character of "OY" at any argument whose date is d (that character

38

is a function from contexts to contents, and the content of "OY" at a context whose date is d will have an extensioncontaining all and only events that occurred on the day before d).

The proof for the second sentence is analogous.

I must mention a possible difficulty. In all my uses of the date name "d" I am assuming that it can pick out thesame date in different possible worlds. I confess that it is very unclear whether this makes sense. Is there anyidentification of dates, or in general times, from one world to another? My intuition is that there is, but I realize thatmany readers, and probably most, will not share this intuition. But for the present purposes I need only somethingweaker and easier to believe. Let us not assume that we can identify dates from one world to another in any absolutesense, but only relative to a fixed point in each world taken as an arbitrary "origin." When we compare pairs ofworlds which are very similar, as we always do in my examples, the choice of origin is natural, since the history ofthe worlds is similar enough to map one frame of reference onto the other. The point is that once a frame of referenceis chosen in each world, it will follow that relative to any context, if an event e occurs in one world on date d, and e'occurs in the other on date d, then either they both occurred yesterday or they both did not occur yesterday.35See Dreier, [6]. The present must be schematic.36We have to distinguish, as Kaplan does, between the necessity of "I am JD," which amounts to the proposition itexpresses being true in every world, and the analyticity of an expression like "I am here now," which amounts to itsbeing true in every context. It is in this sense that "Happiness is good," when uttered in the relevant context, mightbe necessarily true but not analytic.37See Horgan and Timmons, [13], for an elegant argument connecting the problem of reference for moral terms withissues of supervenience.38First posed in Geach, [8]; see also his [9].39For example, in truth functional combinations (where Blackburn might be thought to have made some progress),in propositional attitude attributions, in modal contexts. See Hale, [10].40Perry, [25]. See also Lewis, [20].41See Moore [22] and Blandshard [4].42Moore [22] pp. 33ff for the Contradiction Problem and pp. 42ff for the Disagreement Problem.43Dreier [6].44Blandshard [4] pp. 505-6.45I follow Kaplan at this point, as I do throughout this paper, in holding that indexicals are directly referential, thattheir character determines a different content on different occasions of use, but given an occasion of use, the contentdoes not change according to the character rule when we evaluate what was actually said in other, counterfactualcontexts. I fear this sounds more complicated than it really is; see Kaplan [15].46Railton, [27].47Klagge, [18] and [19].* I thank Jaegwon Kim and Nick Zangwill for helpful conversations about the arguments in this paper, andespecially Ernest Sosa for detailed comments. I also thank an anonymous reviewer for this journal for pointing outtwo problems, which I hope I have corrected.